Characterization of cerebral small vessel disease by neutrophil and platelet activation markers using artificial intelligence

Abstract

Cerebral small vessel disease (cSVD) accounts for 25% of ischemic strokes and is a major cause of cognitive decline. Inflammatory processes, involving immune cells and platelets might drive development and progression of cSVD. The aim of the study was to identify potential novel biomarkers for cSVD, gaining new insights into its pathophysiology. We measured inflammation and platelet and neutrophil activation markers in patients with cSVD and age-matched controls. It was hypothesized that cSVD is accompanied by altered levels of these markers. The levels of interleukin 1β, CX₃CL1, CXCL4, CXCL7, myeloperoxidase (MPO), MPO-DNA complex and S100A8/A9 were measured by ELISA in plasma samples of patients with cSVD presenting with mild vascular cognitive impairment (mVCI, n = 36) or lacunar stroke (Laci, n = 44), and controls (n = 38). To determine the relevance of these ELISA markers compared with patient- and MRI-based characteristics, all characteristics were entered into three machine learning models. Among the ELISA markers measured, MPO levels were significantly elevated in patients with cSVD (48.3 (27.8-80.1, interquartile range) ng/mL) compared with controls (32.2 (19.6-47.4) ng/mL, P = 0.023), particularly in the Laci group (56.8 (33.3-84.7) ng/mL, P = 0.004). Regularized logistic regression and random forest algorithms returned MPO levels as an important feature in the detection and prediction of cSVD. Of note, logistic regression and random forest analysis also highlighted levels of CXCL4, CXCL7, MPO-DNA and S100A8/A9 as features associated with cSVD. Taken together, the neutrophil activation marker MPO is elevated in patients with Laci and machine learning indicates platelet and neutrophil markers as interesting molecules for future investigation. SHORTENED ABSTRACT: Cerebral small vessel disease (cSVD) is a major cause of cognitive decline and stroke. We aimed to identify potential novel biomarkers for cSVD and to obtain new insights into its pathophysiology. Levels of markers reflecting neutrophil activation, neutrophil extracellular trap (NET) formation, platelet activation and vascular inflammation were measured in plasma samples of patients with cSVD, and controls. Only myeloperoxidase (MPO) levels were significantly altered. Regularized logistic regression and random forest algorithms returned MPO levels as an important feature in the detection and prediction of cSVD and highlighted platelet- and NET markers as cSVD associated.

Keywords: Circulating DNA; Inflammation; Lacunar stroke; Machine learning; Neutrophil extracellular trap.